Within the space of this brief review, it's impossible to do more than touch on a few aspects of galaxy interactions; I've said nothing about the tidal genesis of grand-design spirals, origins of ring galaxies, multiple mergers and the formation of cD galaxies, or interactions in groups and clusters, to name a few. The main points I have tried to address are listed here:
1. Simulating interacting galaxies is an art; it can't be reduced to a recipe. Picasso defined art as ``a lie which makes us realize truth'', and this seems to be a good stance to adopt when trying to reproduce real galaxies. Some features of the Antennae may be clues leading to fundamental insights, while others may be due to quirks of the pre-encounter disks. We don't always know which is which; experience is the only guide.
2. In interacting galaxies, thermodynamics is the key to the fate of the gas. Gas which encounters strong radiative shocks in the early phases of a collision will diverge from the stars and fall into the centers of interacting galaxies and merger remnants. Gas which does not suffer such shocks until the later stages of a collision, on the other hand, retains much of its initial angular momentum and builds up extended disks.
3. Remnant structure is determined by many factors. Steep central cusps (or nuclear black holes) may suppress strong triaxiality, but this doesn't explain why galaxies with such profiles rotate rapidly. More generally, sheer existence of self-consistent equilibria is not enough to explain the properties of elliptical galaxies; the details of formation play an important role.
4. Simulations including star formation are still in their early days; a good deal of further work is needed to develop and test alternate approaches. Plausible treatments of feedback from star formation and evolution require abandoning the assumptions which effectively limit the simulated gas to a single phase.
As noted in the abstract, galaxy mergers have deep connections to galaxy formation. For example, the issues reviewed in Sections 2 and 4 arise in cosmological simulations of disk galaxy formation; in dissipative CDM simulations, gas inflows are so efficient that little remains to build disks (Navarro & Benz 1991, Navarro & White 1994, Navarro & Steinmetz 1997). The resolution of this problem is probably to implement strong feedback of the kind long assumed in hierarchical models of galaxy formation (White & Rees 1978, White & Frenk 1991, Kauffmann et al. 1993, Navarro et al. 1995). This may well be the same sort of feedback needed to reproduce the outflows of hot gas in violently interacting starburst galaxies.
Acknowledgements
I thank John Hibbard for allowing me to discuss our unpublished work on NGC 4038 / 9 and Lars Hernquist for providing me with a copy of TREESPH. I'm also grateful to Jun Makino and the University of Tokyo for hospitality while I was writing this report. This research has made use of NASA's Astrophysics Data System Abstract Service.